We now detect errors loading the image and initialise a blank canvas
instead.
This means that it's possible to open HelloThere for editting, and then
change the "type" field to "image/jpeg", and the textarea automatically
switches to a blank bitmap. Next up is making it possible to change the
size of the bitmap.
Get rid of the separate renderContext stack and instead have a parent
pointer on renderer nodes. This lets us walk back up the render tree to
resolve context references